Table 3 Number of transcripts after steps of filtration and time to run ML algorithms on them
| Â | Datasets | ||||||
|---|---|---|---|---|---|---|---|
Known protein biomarkers | All data | ||||||
Steps | All transcripts | Protein coding | Non-coding | All transcripts | Protein coding | Non-coding | |
Number of transcripts after expression filter; biomarkers no filter, all data > 10,000 | 410 | 262 | 149 | 16,173 | 13,688 | 2724 | |
Number of highly correlated features (transcripts); correlation cutoff > 0.75 | 177 | 98 | 37 | 12,047 | 9866 | 1970 | |
Number of transcripts after removing highly correlated features | 234 | 165 | 113 | 4127 | 3823 | 755 | |
Time to run (in seconds) | RF | 10.77 | 8.09 | 6.44 | 196.25 | 169.31 | 32.60 |
NB | 12.34 | 9.38 | 6.63 | 297.81 | 280.27 | 46.05 | |
KNN | 1.03 | 1.10 | 1.11 | 5.63 | 5.62 | 1.78 | |
SVM | 2.25 | 1,07 | 1.05 | 7.51 | 7.48 | 2.72 | |
NNET | 72.37 | 35.84 | 20.12 | 71,044.53 | 56,114.75 | 3125.74 | |